AI agents for small business: start with one job

For a small business, the best use of an AI agent is one agent doing one job well. The jobs that pay back are the repetitive ones that eat the owner's week: answering routine questions, following up new inquiries, booking appointments, sending quotes and handling invoices. Start with the AI features in tools you already pay for, and consider custom work only when a job is specific to your business and costs you ten or more hours a week.

Most guides on this topic are lists of products. This one is about how to decide, because buying five tools is the most common way small businesses waste money on AI.

#The one-job rule

Large companies run AI programs. A small business needs one thing working.

Pick the job using four questions:

1. How many hours a week does it take? Under two, leave it alone.
2. Does it follow rules you could write down? If you could train a new hire to do it from a page of instructions, an agent can probably do it.
3. Is the information it needs in a system? If the answers live only in your head, write them down first.
4. What happens if it gets one wrong? A slightly awkward reply is fine. A wrong quote or a missed booking is not, so those need a person to approve.

The job that scores well on all four is your first agent. Everything else waits.

#Good first jobs, by type of business

For more detail by function, see our guides to agents for customer service, sales, finance, e-commerce and real estate.

#Start with what you already pay for

Before you buy anything, check the tools you have. Most business software has added AI features in the last two years.

• Your email and office suite. Google Workspace and Microsoft 365 both include an AI assistant on business plans. Drafting replies, summarizing threads and pulling answers out of your own documents covers a surprising amount.
• Your booking or scheduling tool. Many now handle rescheduling and reminders by message.
• Your help desk or shared inbox. Most can draft or send replies from your saved answers.
• Your CRM. Lead follow-up, call notes and record updates are common built-in features.
• Your accounting software. Receipt and invoice capture is standard.
• Your online store. Most platforms include an assistant for product questions and order status.

These cost little or nothing extra, need no integration, and keep your data where it already is. They will not be perfect. They will show you what your customers actually ask, which is the most useful thing to know before spending more.

#When it is worth paying for more

Move beyond built-in features when one of these is true.

• The job crosses several tools. An inquiry arrives by email, needs a price from a spreadsheet, a slot from the calendar and a record in the CRM. No single product sees all of it.
• Your rules are your own. How you price, who you serve, what you promise. A generic tool will get these wrong.
• The job takes ten or more hours a week. Below that, the saving rarely covers custom work.
• You have tried the built-in version, and you can say exactly where it falls short.

At that point there are two routes. A no-code agent builder suits an owner or team member who likes to tinker, and a simple job. A custom build suits a job that matters to the business and has to work reliably, because someone else carries the work of connecting systems, testing and handover.

#Realistic budgets

These are approximate ranges drawn from publicly listed prices and typical quotes, not a formal survey. A fuller breakdown is in how much does an AI agent cost.

A quick check on whether custom work pays: multiply the hours the job takes each week by what that time costs you. A job that takes ten hours a week at $40 an hour costs about $1,700 a month. A $6,000 build pays for itself in under four months. The same build for a two-hour job takes well over a year, which is too long. You can run your own numbers in the AI automation ROI calculator.

#Mistakes that waste money

• Buying tools before choosing the job. Five subscriptions, none properly set up.
• Automating a broken process. If quotes are slow because pricing is unclear, an agent will produce unclear quotes faster.
• Letting the agent loose on customers on day one. Start with it drafting and a person sending. Switch to automatic only for the categories where it is reliably right.
• Nobody owns it. Someone needs to read a sample of conversations each week and keep the information current.
• No way to reach a person. Customers forgive an AI. They do not forgive being trapped by one.
• Staff using personal AI accounts for customer data. A one-page policy fixes this. Our AI acceptable use policy template is free to copy.

#A simple plan for the first month

1. Week 1. For one week, note every repetitive task and the minutes it takes. Apply the four questions and pick one job.
2. Week 2. Write down the rules and the answers for that job. Turn on the relevant AI feature in the tool you already have.
3. Week 3. Run it with a person approving everything. Keep a list of what it gets wrong.
4. Week 4. Fix the information behind the mistakes. Let it run on its own for the categories it handles well. Decide whether the built-in feature is enough.

If it is enough, you are done, and it cost you almost nothing. If it is not, you now know exactly what you need, which makes any quote you ask for more accurate and cheaper.

#Where to start

If you want a second opinion on which job to pick, our AI consulting for small business starts at $1,500, and the AI Opportunity Audit is a one-week review at the same price that ends with a ranked list of opportunities and what each would cost. If you already know the job, book a free 20-minute call. We will tell you whether a tool you already own can do it before we suggest building anything.

This guide is part of our series on AI agents for business.

Common questions.

• How are small businesses using AI agents?

  Mostly for a handful of repetitive jobs: answering common customer questions by chat and email, replying to new inquiries within minutes, booking and rescheduling appointments, sending quotes and follow-ups, keeping customer records current, and reading invoices and receipts into the accounts. The businesses that benefit most pick one of these and get it working properly before adding another.

• Which AI agent is best for a small business?

  Usually the one inside a tool you already use. Your booking system, help desk, CRM, accounting software and email suite probably all have AI features now, and they need no integration work. A separate agent tool or a custom build is worth considering only when those cannot do the job, or the job is particular to how your business runs.

• How much does an AI agent cost for a small business?

  Built-in AI features are often included in your plan or cost tens of dollars per user per month. A dedicated agent tool typically costs from about $50 to a few hundred dollars a month. A custom agent for one job usually costs $5,000 to $15,000, and a simpler automation costs less. Add a small monthly amount for AI usage.

• Do I need technical staff to run an AI agent?

  Not for built-in features or most agent tools, which are designed for non-technical users. For a custom agent you need someone to own it: read a sample of its conversations each week, keep the information it uses up to date, and know who to call when something changes. That takes an hour or two a week, not an engineer.

• Is my customers' data safe with an AI agent?

  It depends on the tool and the plan. Use business plans, which generally do not train on your data, and avoid staff using personal AI accounts for customer information. Give an agent access only to what its job needs, and check where the provider stores data if you have customers in the EU or UK. A one-page AI policy for staff covers most of the risk.